Volatile Organic Compounds in the Po Basin. Part A: …stream-ucm.es/PDF PUBLICACIONES/ordonez/Steinbacher_VOCs... · 2016. 2. 25. · (Guerra et al., 1995) including benzene, which

Journal of Atmospheric Chemistry (2005) 51: 271–291DOI: 10.1007/s10874-005-3576-1 C© Springer 2005

Volatile Organic Compounds in the Po Basin.Part A: Anthropogenic VOCs

M. STEINBACHER1,2, J. DOMMEN1, C. ORDONEZ1, S. REIMANN2,F. C. GRUEBLER3, J. STAEHELIN3 and A.S.H. PREVOT1

1Paul Scherrer Institut, Laboratory of Atmospheric Chemistry, CH – 5232 Villigen PSI, Switzerland,e-mail: [email protected] Federal Institute for Material Testing and Research (EMPA), Laboratory for Air Pollution/Environmental Technology, CH – 8600 Dubendorf, Switzerland3Swiss Federal Institute of Technology, Institute for Atmospheric and Climate Science,CH – 8093 Zurich, Switzerland

(Received: 13 May 2004; Accepted: 11 March 2005)

Abstract. Measurements of volatile organic compounds (VOCs) were performed in the Po Basin,northern Italy in early summer 1998 within the PIPAPO project as well as in summer 2002 andautumn 2003 within the FORMAT project. During the three campaigns, trace gases and meteorologicalparameters were measured at a semi-rural station, around 35 km north of the city center of Milan. Lowtoluene and benzene concentrations and lower toluene to benzene ratios on weekends, on Sundays,and in August enabled the identification of a ‘weekend’ and a ‘vacation’ effect when anthropogenicemissions were lower due to less traffic and reduced industrial activities, respectively. Recurrentnighttime cyclohexane peaks suggested a periodical short-term release of cyclohexane close to thesemi-rural sampling site.

A multivariate receptor model analysis resulted in the distinction of different characteristic concen-tration profiles attributed to natural gas, biogenic impact, vehicle exhaust, industrial activities, and asingle cyclohexane source.

Key words: benzene, Greater Milan area, positive matrix factorization, toluene, vacation effect,weekend effect

1. Introduction

Volatile organic compounds (VOCs) play a central role in the photochemistry of theatmospheric boundary layer. In the presence of sunlight and catalyzed by the oxidesof nitrogen (NOx ), VOCs lead to the production of ozone (O3) as already shownin the 1950s (Haagen-Smit, 1952). The VOCs present in the boundary layer areof anthropogenic as well as biogenic origin. On a global scale, biogenic emissionsdominate, but in non-rural areas anthropogenic activities (particularly traffic andindustrial processes) can provide large VOC emission sources that can lead to highhydrocarbon concentrations in the urban boundary layer. Due to their rather highmaximum ozone incremental reactivity (Carter, 1994), the aromatics often affect thetropospheric ozone formation most even when the concentrations of other classes of

272 M. STEINBACHER ET AL.

compounds are higher (Sagebiel et al., 1996; Latella et al., 2005). Incomplete com-bustion of fossil fuels releases mostly hydrocarbons (alkanes, alkenes, aromatics),whereas organic solvents and their vapors also contain considerable amounts ofoxygenated hydrocarbons (Friedrich and Obermeier, 1999). Aromatic compoundsare often used as markers to identify the origin of the air masses (Winkler et al.,2002; Christensen et al., 1999) and are of particular interest due to their largeabundance in motor vehicle emissions, their role in the ozone and aerosol formation(Seinfeld and Pandis, 1998; Kalberer et al., 2004), and their adverse health effects(Guerra et al., 1995) including benzene, which is rated as a human carcinogen.

The Milan area, located in the Po Basin is the most industrialized and denselypopulated region in Northern Italy (Neftel et al., 2002). Among the large urbanareas in Europe, the Greater Milan area and the surrounding Po Basin are know asone of the regions that are most strongly affected by primary (Guerra et al., 1995)and secondary pollutants (Prevot et al., 1997; Thielmann et al., 2001; Dommenet al., 2002). A high-resolution emission inventory for a weekday in May 1998for the Po Basin revealed that stationary sources dominate the VOC emissionswhereas the traffic emissions contribute half as much and biogenic emissions areonly significant during the sunlight hours (Dommen et al., 2003). Besides thelarge emissions in this area, typical meteorological conditions characterized byfrequent calm winds, thermal inversions, and high-pressure systems also favor theaccumulation of atmospheric pollution (Bardeschi et al., 1991).

The presented measurements were part of the two field studies, Pianura PadanaProduzione di Ozono (PIPAPO, ozone production in the Po Basin) and Formalde-hyde as a Tracer of Oxidation in the Troposphere (FORMAT). The field experimentPIPAPO took place from May 06 to July 14, 1998 as a part of the EUROTRAC-2subproject Limitation of Oxidation Production (LOOP) (Neftel et al., 2002). Dur-ing this period, field measurements were performed at more than 30 ground stationsbetween the city of Milan and the Alps north of Milan as well as on-board of aresearch aircraft. The project aim was the research on the evaluation of the tem-poral and spatial extent of the VOC and NOx sensitivity of the ozone formation.Two of the main measurement sites during the PIPAPO campaign were Bresso andVerzago, both located north of downtown Milan. VOC data were available for 56days in Verzago and 42 days in Bresso. The hydrocarbon measurements of PIPAPOwere analyzed by Gruebler (1999).

The main objective of the FORMAT project is to obtain a better knowledge ofthe regional distribution of formaldehyde (HCHO) in the Po Basin. HCHO is animportant intermediate of the VOC oxidation and a source of oxidizing radicals.Measurements were conducted at three ground-based stations and on-board oftwo aircrafts and a microlight research aircraft. Two of the measurement sites(Bresso and Verzago) were identical with the ground stations during PIPAPO 1998.Observations were performed in late summer 2002 (July 22 to September 03) andin autumn 2003 (September 12 to October 6). VOC measurements were availablefor 35 and 17 days in 2002 and 2003, respectively.

VOLATILE ORGANIC COMPOUNDS IN THE PO BASIN 273

In this paper we present, analyze, and compare the available measurementsof individual anthropogenic hydrocarbons focusing on the relation to emissionsources. In the companion paper (Steinbacher et al., 2005), we study biogenicVOCs of these campaigns.

2. Measurement Sites, Techniques, and Meteorological Conditions

2.1. MEASUREMENT SITES

The measurements presented in this work were performed at Verzago, around 35km north of the city center, and at Bresso, approximately 5 km north of downtownMilan. The measurement site at Verzago was situated in a semi-rural environment.The measurement containers were placed at the border of a corn field outside asmall village. Local anthropogenic emissions originated only from a small blindalley passing the measurement containers and a few houses nearby. The nearestmajor road that connects Como and Bergamo was located one kilometer northeastof the measurement site. The surrounding area was patchy with fields, meadows,forests, and small villages. Thus, the local environment was not homogeneous withrespect to biogenic and anthropogenic emissions.

The measurements at Bresso were carried out on a private airfield inside thefreeway belt around Milan, within a densely populated suburb of the city withsome light industry. Road traffic and two close gas stations were the main localanthropogenic emission sources. Vehicle emissions were present at nearly everytime of the day due to heavy traffic on the highway and another main road just 50 mwest of the airfield.

In summer, the topography leads to orographically and thermally induced valleywinds. Consequently, southerly winds prevail during daytime between Milan andthe Alpine foothills. Therefore, Po Basin air is advected towards the Alps, the Milanplume drifts north into the direction of Switzerland and Verzago is often locateddownwind of Milan during daytime in summer (Prevot et al., 1997; Staffelbach etal., 1997; Thielmann et al., 2002).

2.2. TRACE GAS MEASUREMENTS

2.2.1. GC-FID (Airmotec HC1010)

A commercial Airmotec HC1010 gas chromatograph (Konrad and Volz-Thomas,2000) (ChromatoSud, Saint Antoine, France) was used in 1998 (at Bresso andVerzago) (Gruebler, 1999) and 2002 (at Verzago) to measure hydrocarbons betweenC4 and C10. Sample air is pulled through adsorption tubes containing CarbopackB and Carbosieve III. After thermal desorption, the hydrocarbons are cryofocusedusing a fused silica capillary packed with Carbopack B and cooled with CO2.Injection onto the chromatographic column proceeds after fast desorption at 350◦C.The stationary phase of the separation column consists of 2.5% phenyl- and 97.5%


methylpolysiloxan (BGB Analytik AG, Anwil, Switzerland). A flame ionizationdetector (FID) is used for detection and quantification. A measuring cycle con-sisted of 29 min sampling, 1 min transferring the sample to the capillary column,and 9 min for chromatography. A new sample can be drawn while analyzing theprevious sample. Therefore, quasi-continuous measurements with 30 min time res-olution were conducted. In 1998, elaborate calibrations and intercomparisons wereconducted before and after the campaign (Gruebler, 1999). During the campaign,a benzene standard was used for calibration (Thielmann et al., 2001). In 2002,calibrations were performed in the field with a gas standard containing alkanes,alkenes, alkynes, dialkenes, and aromatics ≤C9 at concentrations of a few ppb (Na-tional Physical Laboratory, Teddington, UK) and a benzene standard of 750 ppb(BOC Gase, Multigas, Bienne, Switzerland).

2.2.2. PTR-MS (Ionicon Analytik)

A proton transfer reaction mass spectrometer (PTR-MS) was used for VOC mea-surements at Verzago in 2002. The measuring method is based on a proton-transferreaction of hydronium ions (H3O+) to compounds with a higher proton affinitythan water taking place in a drift-tube reactor, and subsequent detection of theproduct ions in a quadrupole mass spectrometer (Balzers QMG422). Most com-mon volatile organic compounds in the atmosphere, excluding the alkanes andsmall alkenes/alkynes, exhibit sufficient high proton affinities (NIST Chemistrywebbook, 2004). The proton-transfer takes place at relatively low energies andtherefore causes only little fragmentation of most of the ion products. The wholeinlet system both outside and inside the instrument was made out of Silcosteel R©

(Restek, Bellefonte, U.S.A.) tubes. The inlet tubing, as well as the tubing in theinstrument and the drift-tube were temperature stabilized to 50 ◦C to improvethe instrument stability and to minimize the variability of the background signal(Steinbacher et al., 2004). The background signal was measured every 3 hours for30 min using an activated charcoal cartridge in the sampling line. In the defaultmode, 30 selected masses were measured, each of them with a dwell time of 10 s,resulting in a cycle time of 4 1

2 min. Four times a day full scans from 20 amu up to150 amu (dwell time per mass 2 s) were conducted. One full scan cycle took around4 1

2 min. Due to the proton transfer, VOCs are usually identified at their parent mass+ 1. Thus, benzene and toluene were measured at m/z 79 and m/z 93, respectively.However, because of the lack of chromatographic separations, isomers can not bequantified individually. Calibrations were performed with the same standards asused for the Airmotec GC-FID calibration in 2002.

2.2.3. GC-FID (Varian 3400)

For the analysis of VOCs between C2 and C7 in the FORMAT campaign 2003,a gas chromatograph equipped with a flame ionization detector (GC-FID, Varian


3400) was used. VOCs from 200 ml of air were enriched on a cooled microtrap(−30 ◦C) by a thermodesorber (Perkin-Elmer, Turbomatrix) at a flow rate of 10ml/min during 20 min. After trapping, VOCs were introduced directly into the GCby conductive heating of the trap to 250 ◦C at 40 ◦C/min. Analysis was performedon an Al2O3/KCl PLOT column (50 m × 0.53 mm i.d.) with He as carrier gas.Calibration runs were made every week with a 30-component standard in the ppbrange (NPL, Teddington, U.K.).

2.2.4. Other Trace Gas and Aerosol Measurements

Apart from the organic compounds and meteorological parameters, a wide rangeof inorganic compounds (carbon monoxide, ozone, nitrogen oxides, other reac-tive nitrogen, radon) and aerosols were measured during the different campaigns.Species of interest within this work are most of all ozone, nitrogen oxides, andcarbon monoxide. Ozone was measured by UV absorption (Environics) and ni-trogen oxides by ozone chemiluminescence (Eco Physics, MonitorLabs). In 2002,nitrogen dioxide was measured by luminol chemiluminescence (for details seeDommen et al. (2002)). CO was measured at Verzago in 1998 with a conventionalnon-dispersive infrared detector (Horiba), in 2002 and 2003 by UV-vacuum flu-orescence (AeroLaser). Meteorological parameters like temperature, wind speedand direction, humidity, and radiation were measured during the three campaigns.

2.3. METEOROLOGICAL CONDITIONS AND OZONE CONCENTRATIONS

A description of the meteorological conditions during the PIPAPO campaign isgiven by Neftel et al. (2002). Besides the two intensive observation periods (May12–13 and June 1 to 10, 1998) with sunny conditions, a third sunny and warmperiod due to a high pressure system over western Europe occurred from June 16 toJune 23. During the last period, maximum temperatures reached up to 30 ◦C, andozone levels rose to 145 ppb at Verzago whereas maximum O3 mixing ratios of upto 150 ppb and 200 ppb were detected at Bresso and Verzago, respectively, duringthe first intensive period (Spirig et al., 2002; Thielmann et al., 2002).

During FORMAT 2002, three fair weather periods were encountered. The firstperiod appeared from July 22 to July 30, the second from August 6 to 8, and thethird from August 12 to 18. Ozone mixing ratios stayed always below 95 ppb, evenwhen afternoon temperatures reached 30 ◦C, reflecting a generally less pollutedboundary layer.

The FORMAT campaign 2003 started with a foehn event on September 12followed by stable anticylonic conditions associated with a ridge stretched fromEastern Europe towards North Africa at high and mid levels. Meteorologically stableconditions prevailed until September 22 when a trough passed the Po Basin, whichresulted in a drop of the afternoon maximum temperature from 28.2 ◦C (September21) to 21.7 ◦C (September 23). Within the same period, afternoon maximum ozone


mixing ratios dropped from 132 ppb to 44 ppb. The highest ozone mixing ratioswere observed on September 20 with 138 ppb.

3. Results and Discussion

3.1. AROMATIC COMPOUNDS

Figures 1 and 2 show typical time series of aromatic compounds during the cam-paigns in 1998 and 2003. A distinct toluene peak occurred usually around noontime,also accompanied by elevated C2-benzene levels (not shown; usually, the mixingratios of the C2- and C3-benzenes in 1998 stayed below 2 ppb and well below 500ppt, respectively) whereas the benzene concentrations stayed nearly constant. Thepeaks appeared during south-southwesterly winds with wind speeds between 1.5and 2.5 m/s in 1998 and between 1 and 2 m/s in 2003. Several reasons point toan industrial release as the origin of these midday peaks. Firstly, we assume thatindustrial activities take place during the working days but not during weekends.Similar meteorological conditions prevailed on Sunday, 21 June 1998 and Sunday,21 September 2003 but comparable peaks were not detected. Secondly, the useof benzene as a solvent is forbidden in Italy, which could explain the low ben-zene levels even when the other aromatics were high. Thirdly, aromatic emissionsfrom traffic usually show a good correlation between benzene and toluene as itis observed for Bresso in 1998 (see Figure 3, small panel). Figure 3 also showsa characteristic 6-day time series for benzene and toluene at Bresso. Usually, the

Figure 1. Time series of wind speed, ozone, wind direction, and aromatic compounds atVerzago measured with GC-FID in 1998.


Figure 2. Time series of wind speed, ozone, wind direction, benzene, and toluene at Verzagomeasured with GC-FID in 2003.

Figure 3. Big panel: 6-day time series of benzene and toluene at Bresso in 1998. Small panel:scatterplot of toluene and benzene for the whole PIPAPO campaign (May 12 to July 14, 1998).The slope of the orthogonal regression line is 4.4 (R2 = 0.8). Upper panel: time series ofozone.


highest mixing ratios of anthropogenic VOCs were found during the night due toan accumulation of traffic emissions close to the measurement site in the shallownocturnal boundary layer. An extreme event was observed in the early morning of30 May 1998. In contrast to Verzago, the benzene mixing ratios are well correlatedwith the other aromatics even during the day (e.g. see the simultaneous peaks ofbenzene and toluene on 28 May) supporting the hypothesis of traffic emissions asthe main source. Nevertheless, it should be mentioned that the toluene to benzeneratio reflected by a slope higher than 4 in Figure 3 is twice the ratio of the cor-responding emission factors derived from a tunnel study in Switzerland in 1993(Staehelin et al., 1998). Ciccioli et al. (1999) observed in the Greater Rome Areaduring a 2 days case study in 1991 toluene/benzene mixing ratio ratios of 1.9 to2.8 and 1.6 to 1.8 in the urban and in the suburban area, respectively. In Italy, thebenzene content in the gasoline was limited by law to 1% already in 1998 (Dommenet al., 2003; Gruebler, 1999), whereas fuel with up to 5% benzene was allowed inSwitzerland at the same time. The mean benzene content of the Swiss car fleetconstituted 2.3% in 1993 (Heeb et al., 2000). In the beginning of 2000, restric-tions to 1% benzene content were also implemented in Switzerland resulting in atoluene/benzene increase from 1.7 to 3.4 in Duebendorf, a suburban sampling siteclose to Zurich, between 1998 to 2001.

Analyzing the benzene to toluene scatterplot for Verzago in 1998 (see Figure4), a lower correlation than in Bresso is recognizable. Due to the longer lifetime ofbenzene compared to toluene (Seinfeld and Pandis, 1998), it would be expected todetect lower toluene to benzene ratios if the traffic emissions are farther away fromthe measurement site due to the ageing of the air mass. However, higher toluene tobenzene ratios (slope of 7.7) were found around noontime (11 AM to 2 PM) in linewith the interpretation of the advection of industrial emissions at this time. A slopeof 4.7 for all other cases is comparable to the ratio at Bresso and to data obtained fromthe Regione Lombardia (Direzione Generale Qualita dell’Ambiente, http://www.ambiente.regione.lombardia.it/webqa/aria/Archivio.htm) close to roads. We alsodistinguished between weekdays and Saturday and Sunday (Figure 4b). It is obviousthat the lowest toluene to benzene ratios and the highest correlation without anyenhanced toluene mixing ratios were found on Sundays. A slightly higher slopewith a few cases of elevated toluene mixing ratios was measured on Saturdayswhereas the largest scatter and most of the data with high toluene to benzene ratioswere observed on weekdays. Until now, the weekend effect for ozone (partly incombination with other air pollutants like CO and total VOCs) is well documented(Bronnimann and Neu, 1997; Diem, 2000; Jenkin et al., 2002; Marr and Harley,2002; Heuss et al., 2003). However, there are only a few studies on the weekendeffect on certain VOC species. Blanchard and Tanenbaum (2003) showed for 10sites in California that the benzene and toluene levels are considerably reduced onweekends, especially on Sundays. Figure 4b nicely reveals that a ‘weekend effect’in the toluene to benzene ratios can be detected despite the limited number of dataduring a 70-day field campaign.


Figure 4. Scatterplot of toluene vs. benzene at Verzago in 1998. Panel a) distinction of casesaround noontime (11 AM to 2 PM) and other daytimes. The slopes of the orthogonal regressionline are 7.7 and 4.7 for 11 AM to 2 PM data (dashed line, R2 = 0.53) and the rest of the data(solid line, R2 = 0.64), respectively. Panel b) distinction of weekday, Saturday, and Sundaycases. The slopes of the orthogonal regression lines are 5.6 (weekday, grey line, R2 = 0.59),3.8 (Saturday, solid black line, R2 = 0.70), and 3.4 (Sunday, dashed line, R2 = 0.85).

Figure 5a shows a corresponding plot to Figure 1 for mid-August 2002. Incontrast to 1998, the VOCs were measured by PTR-MS. The black horizontal barsstress the periods with the conditions that prevailed in 1998 when toluene peaksaround noontime occurred. It is obvious that on average no similar behavior could bedetected within the presented period except at the end of the 2002 campaign, when


Figure 5. Time series of ozone, m/z79 (protonated benzene) and m/z93 (protonated toluene) atVerzago measured with PTR-MS in 2002. Panel a) August 15 to 21. Panel b) August 25 to 31.

such peaks were observed again as it is seen in Figure 5b (C2-and C3-benzeneswere usually below 500 ppt and 200 ppt, respectively. The sum of C3-benzenesmeasured with the PTR-MS exceeded only rarely 500 ppt). That seems to be aresult of the ‘ferragosto’, the Italian vacation time effect in mid-August, whenindustrial activities and traffic are considerably reduced because by the end ofAugust the vacation time is terminated in this area. This effect is also corroboratedby different slopes of the toluene versus benzene plot that reveals the differentconditions during the 3 weeks in mid-August (see Figure 6). The slopes at thebeginning and at the end of August are nearly identical whereas the slope in betweenis considerably lower. The meteorological conditions were similar during the firstand the second period and no considerably different ozone levels were found.


Figure 6. Toluene to benzene scatterplot for the 2002 campaign. Different periods are charac-terized by different symbols. The slopes of the orthogonal regression line are 5.3 for the first(R2 = 0.36; solid black line) and the third period (R2 = 0.40; dashed black line) and 2.5 forthe second period (R2 = 0.43; grey line).

Therefore a strong bias due to different photochemical oxidation capacities of theatmosphere is not expected. Consequently, a different emission pattern (due todifferent industrial activities) is likely to be responsible for the different tolueneto benzene ratios. The mean toluene/benzene ratios are 3.9, 2.7, and 3.2 for thefirst, the second, and the third period, respectively. Perennial benzene and toluenedata from the Direzione Generale Qualita dell’Ambiente, Regione Lombardia formeasurement sites close to traffic emissions in the Po Basin show considerablyreduced benzene and toluene mixing ratios in August, too, as well as significantchanges (> 99.9%; two-population (independent) t-test) in the toluene to benzeneratio similar to the ones observed at Verzago. The mean afternoon (10 AM to 6 PM)benzene and toluene mixing ratios dropped in Como (Viale Cattaneo), Monza (ViaElvezia), and Milano (Via de Vincenti) from July to August as listed in Table I. AtVerzago, the mean benzene ratios were already low in July and no reduction in themean benzene was observed. The slightly smaller decrease of the ratio at the threetraffic dominated sampling sites compared to our semi-rural site could be explainedby a more dominating industrial impact (including a reduced industrial impact inmid-August) at Verzago. Similar observations with significantly reduced tolueneconcentrations and rather constant benzene levels in August compared to July werealready published for a two years dataset in Rome but were not explicitly relatedto a changed emission pattern during the summer holidays (Brocco et al., 1997).

Table II summarizes a compilation of toluene to benzene ratios at Verzago forthe three campaigns and the two different effects discussed above. On the one hand,the lowest toluene to benzene ratios were always measured on Sundays, and the


Table I. Mean afternoon (10 AM to 6 PM) toluene and benzene mixing ratios (in ppb), andtoluene to benzene ratios for July and August

July August

Toluene/ Toluene/Station Toluene Benzene Benzene Toluene Benzene Benzene

Comoa 8.5 1.9 4.8 6.6 1.5 4.4

Monzaa 7.5 1.5 7.1 4.7 1.1 5.3

Milanoa 3.9 1.1 3.5 2.4 0.9 2.7

Verzago 0.8b 0.3b 3.9b 0.6c 0.3c 2.3c

aObtained from the Direzione Generale Qualita dell’Ambiente, Lombardia.b22 July to 04 August 2002.c05 August to 25 August 2002. Como, Monza, and Milano data were available from 1996–2003,2000–2003, and 2001–2002, respectively.

Table II. Mean (±standard error of the mean) toluene to benzene ratios atVerzago on weekdays, Saturdays, and Sundays for the three different campaigns

Toluene/Benzene ratio

Campaign Weekday Saturday Sunday

PIPAPO 1998 3.77 ± 0.04 2.84 ± 0.07 2.30 ± 0.03

FORMAT 2002, all 3.26 ± 0.11 2.89 ± 0.08 2.57 ± 0.07

FORMAT 2002, 05.08. to 25.08. 2.73 ± 0.06 2.87 ± 0.10 2.50 ± 0.08

FORMAT 2002, rest 3.59 ± 0.18 2.92 ± 0.12 2.66 ± 0.10

FORMAT 2003 3.35 ± 0.11 3.21 ± 0.11 2.13 ± 0.08

The campaign in 2002 is additionally split into two periods to emphasize the‘ferragosto effect’.

highest on weekdays. On the other hand, the toluene to benzene ratios in mid-August 2002 were similar on weekdays and on weekends, confirming the reducedindustrial emissions during that month.

3.2. CYCLOHEXANE

The behavior of cyclohexane suggests also a significant industrial source (Figure 7).The background mixing ratios between 25 and 200 ppt are in a good agreement withdata observed in the suburban area of Rome (Ciccioli et al., 1999). During mostof the nights, distinct cyclohexane peaks up to 9 ppb appeared between 9 and 10PM whereas levels usually below 500 ppt were observed during the rest of the day.No peaks were measured at weekends, which points again to an industrial release.


Figure 7. Time series of cyclohexane at Verzago measured with GC-FID in 2003. Weekendsare marked with ellipses.

Figure 8. Scatterplot of cyclohexane versus wind direction (big panel) and wind speed (smallpanel) for the 2003 field campaign.

Cyclohexane is used for the manufacture of nylon, the production of syntheticfibers and as a solvent in varnishes, resins, and stain removers. The Air ResourcesBoard (http://www.arb.ca.gov) reported primary stationary sources in Californiafrom petroleum refining, automotive repair shops, and commercial printing andpublishing. Figure 8 shows the correlation between the wind conditions and thecyclohexane mixing ratios. It reveals that the point source is located in the north


and the mixing ratios are advected most efficiently with wind speeds between 2and 3 m/s. In contrast to peaks of ethene and propene coincident with the turn ofthe wind direction from west to north (see the companion paper (Steinbacher et al.,2005)), the cyclohexane peaks appeared around 1 to 2 h after the change of the winddirection showing that sources for cyclohexane and the alkenes were different.Like the source of the alkenes, the emission source of cylohexane is also located inthe north. The short and narrow peaks of cyclohexane suggest a point source withina couple of kilometers, most probably between the sampling site and the Alpinefoothills in the north. Assuming a wind speed of 1.5 m/s and considering the time of1 to 2 h after the wind turn, the maximum distance is 10 km. It points to a periodicalshort-term release of cyclohexane possibly at the end of a production process.

3.3. RECEPTOR MODELING USING POSITIVE MATRIX FACTORIZATION ANALYSIS

A multivariate receptor model was employed for air pollution concentration appor-tionment. Such statistical analysis allows attributing the variabilities of trace gasmeasurements to different concentration profiles. The commonly used method ofChemical Mass Balance models needs a priori knowledge of the individual sourceprofiles, which is often not available (Watson et al., 2001). Here we use so called‘Positive Matrix Factorization’ (PMF), a variant of factor analysis with non-negativefactor elements (Paatero and Tapper, 1994; Paatero, 1997) in order to analyze themeasurements of 2003. The results indicate that five different profiles were requiredto fit the field measurements to identify and distinguish various striking concen-tration profiles (see Figure 9). Concentration profile #5 is obviously determinedby the cyclohexane concentration. This profile does not contain any substantialcontribution of another species, revealing that only cyclohexane is emitted by thiscertain industrial process. The diurnal pattern of the contribution to profile #5 (seeFigure 10) nicely represents the diurnal cyclohexane pattern. Concentration profile#2 is the only profile with a substantial isoprene concentration. As the diurnal cycleof the contribution to profile #2 (see Figure 10) looks quite similar to the real iso-prene cycle (see the companion paper (Steinbacher et al., 2005)), this profile canbe attributed to a biogenic impact. Profile #3 is identified as the vehicle exhaust(tailpipe) concentration profile. It shows a toluene to benzene ratio of 2.5 (in goodagreement with the Sunday toluene/benzene traffic ratio, see Table II), and a highethane, ethyne, and propene concentration that is characteristic for vehicle exhaustemissions (Borbon et al., 2003). In addition, our attribution is also confirmed bystudies from the United States performed in road tunnels (Sagebiel et al., 1996),receptor model and emissions inventory source apportionment studies within theSan Joaquin Valley Air Quality Study/Atmospheric Utilities Signatures, Predic-tions, and Experiments Regional Model Adaptation Program (SARMAP) (Fujitaet al., 1995), the Paso del Norte Ozone Study (Fujita 2001), and source apportion-ment studies by chemical mass balance (Watson et al., 2001). The lower toluene tobenzene ratios compared to the data observed at Bresso reveal that photochemically


Figure 9. Concentrations profiles of the positive matrix factorization analysis applied to theVOC dataset measured in 2003. Methylpentane represents the sum of 2-methylpentane and3-methylpentane.

Figure 10. Mean diurnal cycle of the profile contributions for the period from September 17to 23, 2003.

aged air masses are predominantly analyzed at Verzago. Furthermore, the profilecontribution is significantly correlated with carbon monoxide (CO) (R = 0.73,n = 315), a good marker for traffic emissions that was not included in the PMFanalysis. The profiles #1 and #4 only differ by the high toluene contribution in#4 and the different diurnal pattern. Profile #1 decreases in the afternoon whileprofile 4 increases. Therefore we attribute profile #1 to a constant emission source


whereby the diurnal variation reflects the inverse daily cycle of the mixing height.The composition of profile #1 is indicative of natural gas leakages as it was alsoshown by studies in the United Kingdom and the United States (Derwent et al.,1995; Derwent et al., 2000; Choi and Ehrmann, 2004). The high toluene concen-tration in profile #4 and the apparently enhanced emissions in the afternoon pointto industrial activities. Since also evaporative losses from gasoline might be higherin the afternoon due to higher temperatures, we interpret this profile as a mixture

Figure 11. Diurnal cycles of the profile contributions #1 to #5 and NOx during September2003 at Verzago. ‘Weekday 17 to 23 September’ reflects the data from Wednesday 17 to Friday19 and Monday 22 to Tuesday 23. ‘Sunday’ represents the data on 21 and 28 September.Vertical bars represent the standard deviations of the mean. Due to the small number of data,no variability is shown for Sunday data.


of industrial activities (solvent use, combustion) and partly non-tailpipe emissions(evaporation) of gasoline. In summary, the 5 different profiles could be reasonablyrelated to different processes: profile #1 (natural gas), profile #2 (biogenic impact),profile #3 (vehicle exhaust), profile #4 (industry; solvent use and combustion, non-tailpipe evaporation), profile #5 (cyclohexane).

Figure 11 shows the diurnal cycles of the profile contributions and NOx , sepa-rated into weekday and Sunday data. For the weekday data, average diurnal cycleswere calculated for the period from September 17 to 23 as the meteorological con-ditions for this period were similar to the meteorological conditions on the twoSundays. Substantial differences occurred for profile #5 as the cyclohexane peakswere only detected on weekdays and less significantly for profile #4 during the day-time hours as the industrial influence ceased on Sundays. For profile #3 and NOx

(a marker for combustion processes, especially diesel vehicles) a considerable dif-ference was found in the morning hours whereas no clear difference appeared inthe evening hours for the two selected periods. This agrees with the analyses ofBlanchard and Tanenbaum (2003) for NOx , benzene, and toluene mixing ratios onSundays and weekdays. Our findings are also in line with traffic statistics in Milanfor summer months (Agenzia Regionale per la Protezione dell’Ambiente della Lom-bardia Dipartimento, Milan), which show that for four counting locations there areon average 3.3 times more cars (<7 m length) on weekdays than on Sundays in themorning, but only 1.4 times more in the evening. The number of trucks and busses(>14 m) is even reduced a factor of 5.5 in the morning and 2.4 in the evening.This implies that the various weekday to Sunday differences in both the profilecontribution #3 and the NOx mixing ratios are driven by different traffic densities.

4. Conclusions

Three measurement campaigns within a period of 5 years were performed in the PoBasin in early summer 1998, summer 2002, and autumn 2003. Measurements ofaromatic compounds at Verzago showed a recurrent feature with enhanced toluenepeaks without enhanced concentrations of benzene at around noontime. Closely cor-related benzene to toluene ratios in car exhausts and the absence of toluene peaksduring weekends and the holiday period in summer pointed to an advection of indus-trially released toluene. Different toluene to benzene ratios on weekends comparedto weekdays as well as in August compared to July and September enabled us toidentify a ‘weekend effect’ and a ‘vacation (ferragosto) effect’, respectively.

Cyclohexane measurements revealed another anthropogenic impact at the semi-rural sampling site. Distinct peaks appeared between 9 and 10 PM under northerlywinds and wind speeds between 2 and 3 m/s. No peaks were observed on Saturdaysand Sundays.

A multivariate receptor model analysis for the semi-rural site in 2003 distin-guished five different concentration profiles that could be reasonably attributedto natural gas, biogenic impact, vehicle exhaust, industrial activities, and a single


cyclohexane source. Diurnal cycles of the profile contribution showed character-istic differences on weekdays and Sundays. The vehicle exhaust profile contribu-tion was in line with traffic statistics and could be reasonably explained by trafficdensities.

The selected VOCs presented here and in a companion paper (Steinbacher et al.,2005) illustrated the anthropogenic and biogenic impact on the ambient trace gaslevels in the Po Basin. These detailed analyses will provide a good basis withinthe FORMAT project to investigate the production of formaldehyde from differentVOCs and the role of formaldehyde as an intermediate of the VOC oxidation inthe boundary layer. By means of the present VOC patterns, it should be possibleto elucidate the contributions to the HCHO formation of certain VOCs, to specifythe main HCHO precursors, and to distinguish and to quantify the different forma-tion mechanisms of formaldehyde: primary emissions, secondary formation fromanthropogenic and secondary formation from biogenic precursors.

Acknowledgments

We thank the PIPAPO and FORMAT crews, the Direzione Generale Qualitadell’Ambiente of the Regione Lombardia and the Agenzia Regionale per la Pro-tezione dell’Ambiente della Lombardia Dipartimento, sub-provinciale Citta di Mi-lano for providing data. This work was supported by the Swiss National ScienceFoundation under Grant No. 21-61573.00. The FORMAT project was funded bythe Research Directorate of the European Commission under Grant No. EVK2-CT-2001-00120 and by the Federal Office for Education and Science in Switzerland,BBW nr. 01.0233.

References

Bardeschi, A., Colucci, A., Gianelle, V., Gnagnetti, M., Tamponi, M., and Tebaldi, G., 1991: Anal-ysis of the impact on air quality of motor vehicle traffic in the Milan urban area, AtmosphericEnvironment 25B(3), 415–428.

Blanchard, C. L. and Tanenbaum, S. J., 2003: Differences between weekday and weekend air pollutantlevels in Southern California, Journal of the Air & Waste Management Association 53, 816–828.

Borbon, A., Fontaine, H., Locoge, N., Veillerot, M., and Galloo, J. C., 2003: Developing receptor-oriented methods for non-methane hydrocarbon characterization in urban air – part I: sourceidentification, Atmospheric Environment 37, 4051–4064.

Brocco, D., Fratarcangeli, R., Lepore, L., Petricca, M., and Ventrone, I., 1997: Determina-tion of aromatic hydrocarbons in urban air of Rome, Atmospheric Environment 31(4), 557–566.

Bronnimann, S. and Neu, U., 1997: Weekend-weekday Differences of Near-Surface Ozone Concen-trations in Switzerland for Different Meteorological Conditions, Atmospheric Environment 31(8),1127–1135.

Carter, W. P. L., 1994: Development of ozone reactivity scales for volatile organic compounds, Journalof the Air & Waste Management Association 44, 881–899.


Ciccioli, P., Brancaleoni, E., and Frattoni, M., 1999: Reactive Hydrocarbons in the Atmosphere atUrban and Regional Scales, in Reactive Hydrocarbons in the Atmosphere, edited by C. N. Hewitt,pp. 159–207, Academic Press, San Diego.

Choi, Y.-J. and Ehrmann, S. H., 2004: Investigation of sources of volatile organic carbon in theBaltimore area using highly time-resolved measurements, Atmospheric Environment 38, 775–791.

Christensen, C. S., Skov, H., and Palmgren, F., 1999: C5-C8 non-methane hydrocarbon measure-ments in Copenhagen: concentrations, sources and emission estimates, The Science of the TotalEnvironment 236, 163–171.

Derwent, R. G., Davies, T. J., Delaney, M., Dollard, G. J., Field, R. A., Dumitrean, P., Nason, P. D.,Jones, B. M. R., and Pepler, S. A., 2000: Analysis and interpretation of the continuous hourlymonitoring data for 26 C2-C8 hydrocarbons at 12 United Kingdom sites during 1996, AtmosphericEnvironment 34, 297–312.

Derwent, R. G., Middleton, D. R., Field, R. A., Goldstone, M. E., Lester, J. N., and Perry, R.,1995: Analysis and interpretation of air quality data from an urban roadside location in centrallondon from the period from July 1991 to July 1992, Atmospheric Environment 29(8), 923–946.

Diem, J. E., 2000: Comparisons of weekday-weekend ozone: Importance of biogenic volatile or-ganic compound emissions in the semi-arid southwest USA, Atmospheric Environment 34, 3445–3451.

Dommen, J., Prevot, A. S. H., Baertsch-Ritter, N., Maffeis, G., Longoni, M. G., Gruebler, F. C., andThielmann, A., 2003: High-resolution emission inventory of the Lombardy region: developmentand comparison with measurements, Atmospheric Environment 37, 4149–4161.

Dommen, J., Prevot, A. S. H., Neininger, B., and Baumle, M., 2002: Characterization of the pho-tooxidant formation in the metropolitan area of Milan from aircraft measurements, Journal ofGeophysical Research 107(D22), art.-nr.: 8197, 10.1029/2000JD000283.

Friedrich, R. and Obermeier, A., 1999: Anthropogenic Emissions of Volatile Organic Compounds. InC. N. Hewitt, Reactive Hydrocarbons in the Atmosphere, Academic Press, San Diego, pp. 2–39.

Fujita, E. M., 2001: Hydrocarbon source apportionment for the 1996 Paso del Norte Ozone Study,The Science of the Total Environment 276, 171–184.

Fujita, E. M., Watson, J. G., Chow, J. C., and Magliano, K. L., 1995: Receptor Model and EmissionsInventory Source Apportionments of Nonmethane Organic Gases in California’s San JoaquinValley and San Francisco Bay Area, Atmospheric Environment 29(21), 3019–3035.

Gruebler, F. C., 1999: Reactive hydrocarbons in the Milan area: Results from the PIPAPO campaign.PhD Thesis. Institute for Atmospheric Science, Swiss Federal Institute of Technology Zurich.

Guerra, G., Iemma, A., Lerda, A., Martines, C., Salvi, G., and Tamponi, M., 1995: Benzene emissionsfrom motor vehicle traffic in the urban area of Milan: Hypothesis of health impact assessment,Atmospheric Environment 29(23), 3559–3569.

Haagen-Smit, A. J., 1952: Chemistry and Physiology of Los Angeles Smog, Industrial and Engineer-ing Chemistry 44(6), 1342–1346.

Heeb, N. V., Forss, A.-M., Bach, C., Reimann, S., Herzog, A., and Jackle, H. W., 2000: A comparisonof benzene, toluene and C2-benzenes mixing ratios in automotive exhaust and in the suburbanatmosphere during the introduction of catalytic converter technology to the Swiss Car Fleet,Atmospheric Environment 34, 3103–3116.

Heuss, J. M., Kahlbaun, D. F., and Wolff, G. T., 2003: Weekday/Weekend Ozone Differences: Whatcan we learn from Them?, Journal of the Air & Waste Management Association 53, 772–788.

Jenkin, M. E., Davies, T. J., and Stedman, J. R., 2002: The origin and day-of-week dependence ofphotochemical ozone episodes in the UK, Atmospheric Environment 36, 999–1012.

Kalberer, M., Paulsen, D., Sax, M., Steinbacher, M., Dommen, J., Prevot, A. S. H., Fisseha, R.,Frankevich, V., Zenobi, R., and Baltensperger, U., 2004: Identification of polymers as majorcomponents of atmospheric organic aerosols, Science 303, 1659–1662.


Konrad, S. and Volz-Thomas, A., 2000: Characterization of a commercial gas chromatography-flameionization detection system for the in situ determination of C5-C10 hydrocarbons in ambient air,Journal of Chromatography A 878, 215–234.

Latella, A., Stani, G., Cobelli, L., Duane, M., Junninen, H., Astorga, C., and Larsen, B. R., 2005: Semi-continuous GC Analysis and Receptor Modelling for Source Apportionment of Ozone PrecursorHydrocarbons in Bresso, Milan 2003, Journal of Chromatography A 1071, 29–39.

Marr, L. C. and Harley, R. A., 2002: Spectral analysis of weekday-weekend differences in ambientozone, nitrogen oxide, and non-methane hydrocarbon time series in California, AtmosphericEnvironment 36, 2327-2335.

Neftel, A., Spirig, C., Prevot, A. S. H., Furger, M., Stutz, J., Vogel, B., and Hjorth, J., 2002: Sensitivityof photooxidant production in the Milan Basin: An overview of results from a EUROTRAC-2Limitation of Oxidant Production field experiment, Journal of Geophysical Research 107(D22),art.-nr.: 8188, 10.1029/2001JD001263.

NIST Chemistry webbook, 2004: http://webbook.nist.gov/chemistry/pa-ser.html.Paatero, P., 1997: Least squares formulation of robust non-negative factor analysis, Chemometrics

and Intelligent Laboratory Systems 37, 23–35.Paatero, P. and Tapper, U., 1994: Positive matrix factorization: A non-negative factor model with

optimal utilization of error estimates of data values, Environmetrics 5, 111–126.Prevot, A. S. H., Staehelin, J., Kok, G. L., Schillawski, R. D., Neininger, B., Staffelbach, T., Neftel,

A., Wernli, H., and Dommen, J., 1997: The Milan photooxidant plume, Journal of GeophysicalResearch 102(D19), 23375–23388.

Seinfeld, J. H. and Pandis, S. N., 1998: Atmospheric Chemistry and Physics: From Air Pollution toClimate Change, John Wiley & Sons, Inc., New York, pp. 1326.

Sagebiel, J. C., Zielinska, B., Pierson, W. R., and Gertler, A. W., 1996: Real-world emissions andcalculated reactivities of organic species from motor vehicles, Atmospheric Environment 30(12),2287–2296.

Spirig, C., Neftel, A., Kleinman, L. I., and Hjorth, J., 2002: NOx versus VOC limitation of O3 pro-duction in the Po valley: local and integrated view based on observations, Journal of GeophysicalResearch 107(D22), art.-nr.: 8191, 10.1029/2001JD000561.

Staehelin, J., Keller, C., Stahel, W., Schlapfer, K., and Wunderli, S., 1998: Emission Factors fromRoad Traffic from a Tunnel Study (Gubrist Tunnel, Switzerland). Part III: Results of OrganicCompounds, SO2 and Speciation of Organic Exhaust Emissions, Atmospheric Environment 32(6),999–1009.

Staffelbach, T., Neftel, A., Blatter, A., Gut, A., Fahrni, M., Staehelin, J., Prevot, A. S. H., Hering, A.M., Lehning, M., Neininger, B., Baumle, M., Kok, G. L., Dommen, J., Hutterli, M., and Anklin,M., 1997: Photochemical oxidant formation over southern Switzerland. 1. Results from summer1994, Journal of Geophysical Research 102(D19), 23345–23362.

Steinbacher, M., Dommen, J., Ordonez, C., Reimann, S., Staehelin, J., Andreani-Aksoyoglu, S., andPrevot, A. S. H., 2005: Volatile organic compounds in the Po Basin. Part B: biogenic VOCs,Journal of Atmospheric Chemistry 51, 293–315.

Steinbacher, M., Dommen, J., Ammann, C., Spirig, C., Neftel, A., and Prevot, A. S. H., 2004:Performance characteristics of a proton-transfer-reaction mass spectrometer (PTR-MS) derivedfrom laboratory and field measurements, International Journal of Mass Spectrometry 239, 117–128.

Thielmann, A., Prevot, A. S. H., Gruebler, F. C., and Staehelin, J., 2001: Empirical ozone isoplethsas a tool to identify ozone production regimes, Geophysical Research Letters 28(12), 2369–2372.

Thielmann, A., Prevot, A. S. H., and Staehelin, J., 2002: Sensitivity of ozone production derived fromfield measurements in the Italian Po basin, Journal of Geophysical Research 107(D22), art.-nr.:8194, 10.1029/2000JD000119.


Watson, J. G., Chow, J. C., and Fujita, E. M., 2001: Review of volatile organic compound sourceapportionment by chemical mass balance, Atmospheric Environment 35, 1567–1584.

Winkler, J., Blank, P., Glaser, K., Gomes, J. A. G., Habram, M., Jambert, C., Jaeschke, W., Konrad, S.,Kurtenbach, R., Lenschow, P., Lorzer, J. C., Perros, P. E., Pesch, M., Prumke, H. J., Rappengluck,B., Schmitz, T., Slemr, F., Volz-Thomas, A., and Wickert, B., 2002: Ground-based and airbornemeasurements of nonmethane hydrocarbons in BERLIOZ: Analysis and selected results, Journalof Atmospheric Chemistry 42, 465–492.

Documents

Volatile Organic Compounds in the Po Basin. Part A: …stream-ucm.es/PDF PUBLICACIONES/ordonez/Steinbacher_VOCs... · 2016. 2. 25. · (Guerra et al., 1995) including benzene, which